The interaction of images and text during comprehension of garden-path sentences: is integration better than good enough?

One of the main goals of reading is to construct a discourse representation. However, when information in the discourse is confusing or ambiguous, readers are often not able to create a fluid, accurate understanding of the text. Decades of research on temporarily ambiguous garden-path sentences have given the field of psycholinguistics a comprehensive understanding of how readers process, and often misinterpret, garden-path sentences. Both online and offline data demonstrate that when reading garden-path sentences (e.g., As the guard and officer wrestled the thief that was fleeing fell down the stairs), readers may not arrive at the correct, syntactically-licensed interpretation (i.e., that the guard and officer are wrestling each other) despite apparent reanalysis. The garden-path sentence processing model claims that readers often parse sentences by developing only shallow representations of the structure, and thus never return from "down the garden path" to arrive at the correct, syntactically licensed interpretation (Ferreira, Ferraro, & Bailey, 2002; Ferreira & Patson, 2007). In this dissertation, I seek to test the bounds of the garden-path model of sentence processing by investigating contextual influences on language processing. Specifically, how are reading processes are affected by the presence of extralinguistic information (e.g., imagery) with garden-path sentences. If readers are able to properly make use of linguistic (words) and nonlinguistic (pictorial) information, does it help them avoid inaccurate interpretation? For means of comparison, a related question is the role of imagery on linguistic disambiguation for less proficient readers who might rely more on the nonlinguistic code (e.g., nonnative speakers). I approach the issue of visual and linguistic (multimedia) processing by investigating specifically whether or not nonnative speakers of English, compared to native speakers, rely differently on non-linguistic information to parse and understand ambiguous, garden-path sentences